HCP portals in the AI era: getting cited, staying compliant

Healthcare professionals are changing how they find information. Many now ask an AI assistant first and visit a brand's portal only when they need to check a detail, go deeper, or use a specific tool. For pharma teams, that raises three questions at once. How do you get AI to cite your content? How do you publish faster when every page goes through MLR review? And how do you do both without putting compliance at risk?

We put those questions to two agency leaders who build and run HCP portals every day: Alex Jijie, CEO of Digitalya, and Jeremy Summers, CTO of Crescendo Collective. Mark Melton, US Strategy & Partner Lead at Magnolia DXP, moderated the session. Here's what they shared, and what it means for your portal.

Traffic to HCP portals is falling, and to Jijie that isn't necessarily bad news. It reflects a wider shift toward zero-click search, where the answer shows up before anyone visits a site. "I don't remember the last time when I Googled something," he said, "or if I Google, I usually use the AI mode. The HCPs are also doing the same." In his experience, HCPs now use AI to get a preview of the answer. Then they come to the portal for a specific task, like double-checking a piece of information, finding more detail, or using a video or engagement tool. He cited industry research estimating that about three in four HCPs already use AI at work.

A common fear sits behind this trend: that AI will make portal content irrelevant. Summers turns that around. "It doesn't kill your content, it actually consumes it," he said. "HCP portals kind of need to turn into a data warehouse that can feed the AI." The job of a portal changes from being a destination to being the source that AI answers are built on.

Getting cited in AI answers is known as Generative Engine Optimization (GEO). Summers breaks it into three lanes:

A GEO audit adds a content layer on top of the technical checks. Beyond schema and structure, Summers said a good audit shows where your portal leaves questions unanswered, or where the answers live on someone else's site. It also forces an honest question: is this page brochureware, or does it help an HCP do their job?

AI makes it easy to produce more content. In pharma, that's the easy part. Every page still has to pass MLR review, and Summers has worked on projects where MLR is a dedicated, budgeted phase of the roadmap. With AI in the mix, teams could push many times more content into the same review pipeline.

"When we're scaling up here, step one isn't let's make more content. It's let's expand this road," Summers said. He shared a line someone once gave him that keeps him focused: "There's no ROI on content until it's published."

Summers sees AI showing up at both ends of the pipeline. At the start, it helps teams draft and adapt content. At the end, review tools now include AI agents that run a first quality check before human reviewers step in. The gap is in the middle, especially for digital content. Every page needs screenshots, often one for each tab and carousel state. Then someone has to annotate them by hand, drawing boxes on a PDF to explain what each element does and where each link goes. A single web page can turn into a 15- to 20-page submission.

That prep work is the most labor-intensive step, and it's where scaled-up content breaks first. Review processes are often thin compared to the marketing effort around them, and they rarely have a single owner. As demand grows, review timelines stretch without anyone noticing. It's something that "quietly breaks," in Summers's words. Clean, well-prepared submissions also help the AI review tools do their job.

Jijie shared a small example with a large cost. On one patient portal, a single dollar sign in a patient support program animation held up a new page for a week. That kind of issue is exactly what an AI pre-check can flag before a page goes to review. But he was clear about who makes the call: "You always keep human in the loop." AI can surface issues. People decide what gets approved and what goes back for changes.

Jijie recalled a warning from a former AstraZeneca leader: AI has made it very easy for marketers to "scale mediocrity." The smarter move is often to reuse what already works. In one example he shared, a pharma company took existing webinars with key opinion leaders (KOLs) and used AI tools to turn each recording into a transcript PDF, a podcast for HCPs to listen to on the way to work, and short, bite-sized clips placed at different points in the portal. It was the same content, repurposed. According to Jijie, it lifted portal engagement by 13%. Modular content takes the same idea further: teams build pages from MLR-approved modules they can reuse across channels and markets. Mark Melton explains how in Powering modular approaches with a DXP in the biopharma industry.

Company size changes the answer too. In large pharma companies, Jijie sees the Q4 crunch overload MLR teams and medical science liaisons (MSLs) as everyone pushes projects through after the summer. There, better planning may help more than AI. A smaller biotech with only two MSLs has a different problem, and using AI wisely to scale non-scientific content can make a real difference.

Asked what should never go through generative AI, the panel agreed on the answer. Jijie was direct: "Don't let them generate scientific content," meaning the data collected in clinical trials. Summers extended that to claims. If you use AI for content, your approved data, science, and claims should be carried over word for word, not rewritten. Melton added important safety information to the list. Those are the parts of a page where human authorship and review are not optional.

When fewer HCPs visit, each visit counts for more. "Personalization right now, it's becoming a bit more important than it used to be in the past," Jijie said, "because you can really feed right now relevant information to the HCPs out there." He described three levels teams use today:

There's a threshold, though. Personalization pays off on portals with a real content library, around 20 pages or more. On a five-page site, there's not enough to personalize.

Summers made the case in simple terms: "If people don't find what they're looking for, they leave." Many HCP portals already require a login, so teams know the visitor's role and specialty. That makes even basic personalization easy, and in his words, "there's no reason not to."

Europe adds its own challenge. In most European countries, HCP portals need a login, and many registration forms still ask for 20 fields up front without saying what the HCP gets in return. Jijie's advice is to shorten the form and state the value next to it, such as medical information or KOL interviews that aren't available publicly.

Summers's team worked through this country by country with compliance teams, asking one question: what do we actually need up front to meet our legal obligations? The answer varied by market. In many cases, though, they cut the initial ask and moved to progressive profiling, which gathers more detail over time instead of all at once.

Timing matters as much as form length. Jijie pointed to rare diseases, where a condition may be treated by only 50 HCPs across Europe. It's tempting to skip the portal because the field team can call each of them. But an HCP looking for an answer at 11 at night can't wait two weeks for an account approval. By then, the moment and often the brand are forgotten. Where HCPs are already in your CRM with validated consent, integrating registration with it removes that wait. A single sign-in across your brand sites saves them from juggling passwords.

AI models can't see content behind a login, and that's by design. The downside is that gated content is invisible to AI search, SEO, and first-time visitors alike. Summers's approach is to agree with compliance and legal on a public teaser for each gated area. It gives AI and search engines something to pick up and shows visitors what they get after registering. Looking further ahead, the panel discussed on-site AI assistants that can draw on gated content for logged-in users. Any assistant like that should tell users clearly that they're talking to AI. That is a transparency requirement the EU AI Act sets for chatbots from August 2026.

Five years out, neither speaker sees HCP portals going away. Summers expects portals to matter even more as the source of truth AI tools draw from, and as the place experts go when an AI answer doesn't sound right. "Let me see your references," as he put it. Jijie expects portals to become a center of the customer experience, bringing together CRM data, field visits, and external signals.

The work to get there resolves a familiar trade-off. Marketing teams need to publish faster, with structured, personalized content that AI can cite. Medical, legal, and IT teams need governance, traceability, and human approval at every step. Magnolia DXP, the Agentic Digital Experience Platform (DXP), is built for that balance: composable content orchestration for marketing teams, with the governance and human-in-the-loop control that regulated industries need. It's backed by more than 20 years in the market, and Magnolia DXP is ISO 27001:2022 and ENS (High) certified and SOC 2 Type II attested, with GDPR and DORA readiness.